Poke most parts of the chip business and it won’t take long before you run into Arm. The Cambridge, UK-based company sells blueprints needed to design microprocessors and licenses technology known as instruction sets that dictate how software programs communicate with those chips. The technology has been used in more than 250 billion semiconductors since Arm was founded in 1990; it now goes into just about every smartphone on the planet, as well as Apple Inc’s Mac computers and servers at Amazon.com Inc’s data centers.
Arm’s importance to the industry has also been underscored recently by revelations, reported by Bloomberg News and other outlets, that it’s been engaged in pre-IPO talks with a handful of big industry names, including Nvidia Corp and Intel Corp, to become so-called anchor investors. The initial public offering Arm is planning for the fall could be the year’s biggest; it could also provide an interesting temperature check on the fever over artificial intelligence.
Going public could be a welcome validation for Arm, which has been on quite a ride. It traded publicly until 2016, when Japanese investment firm SoftBank Group Corp bought it for US$32 billion. At the time, SoftBank said its ownership would allow Arm to pursue growth without the constant pressure to impress Wall Street with quarterly earnings. Arm grew by about 2,000 people and aggressively pursued new chip designs, while SoftBank founder Masayoshi Son called the company “the center of the center of SoftBank.”
In 2020, SoftBank announced it had reached a deal to sell Arm to Nvidia for US$40 billion. The move was timed to take advantage of a booming semiconductor industry and was also part of SoftBank’s broader selloff of valuable assets. But the idea of Arm’s technology falling into Nvidia’s hands alarmed industry rivals. Executives at some of those companies even suggested establishing a consortium that would buy a chunk of Arm to keep any one entity from ever gaining full control over technologies that had become industry standards.
The ensuing regulatory pressure spurred by industry complaints forced Nvidia to abandon its bid in early 2022. As the deal died, the fortunes of both Arm and SoftBank worsened. Arm Chief Executive Officer Simon Segars, an early employee who’d held the job since 2013, resigned. In China, Arm was involved in an odd power struggle: The CEO of its Chinese subsidiary, Allen Wu, was running day-to-day operations over the objections of Arm’s directors, who’d officially fired him in 2020 over what they said were his conflicts of interest. For its part, SoftBank needed to realize a big return on its Arm investment to avoid the brewing disaster stemming from the diminished value of its portfolio of billion-dollar startups. Once the Nvidia deal fell through, Son announced plans to take Arm public in what he said he hoped would be the “biggest IPO in semiconductor history.”
It was a terrible time to be peddling a new chip stock. The pandemic’s surge in demand for personal computers and smartphones, the two biggest markets for chips, had petered out, leaving device makers with unused stockpiles. Semiconductor companies that had been dealing with chip shortages were left with gluts, and the benchmark Philadelphia Stock Exchange Semiconductor Index lost a third of its value in 2022. SoftBank decided to postpone the IPO and wait for a more opportune window.
See also: Testing for Resource library
That time seems to have arrived. Arm finally forced Wu out, clearing the way to resume normal activity in the largest market for chips. (Wu did not respond to requests for comment.) Also, the launch of ChatGPT in November 2022 sent investors flocking to companies that seemed positioned to capitalize on an AI economy. The performance of Arm’s IPO will depend in large part on how well it can fit itself into that story.
CEO Rene Haas, who took over from Segars last year, has focused on reducing Arm’s reliance on the stagnating smartphone business by pursuing customers that build chips for data centers. Processors for that market are among the most expensive the industry creates and the most profitable. The equipment has also been closely associated with emerging AI products.
Arm sees an opportunity to promote its specific expertise to companies making data center chips, because it’s developed its technology for the energy-constrained world of battery-powered phones. While energy use had generally been less of a concern at data centers, the immense amount of electricity the largest data centers now consume is making efficiency a top priority.
See also: Testing for Resource library
Oracle founder Larry Ellison said at an event on June 28 that more efficient chips are essential for the type of data centers needed for the AI era. Some racks in Oracle facilities remain half full, he said, because Oracle can’t get enough electricity. Ellison is backing a startup called Ampere Computing LLC, which is using Arm’s technology to build processors that replace those from Intel and Advanced Micro Devices Inc. Amazon Web Services is further along than Oracle in overhauling the architecture of the chips it uses in its data centers and has already installed multiple generations of its own Graviton processors, which are also based on Arm’s technology. Amazon says its Arm-based chips are more efficient both in terms of energy and economics, and are being used by 40,000 of its AWS customers.
In its most recent financial disclosures, covering the fiscal year that ended in March, Arm said it increased its revenue by 5.7% compared to the previous fiscal year. Showing such sales growth in the face of a contracting smartphone industry points to Arm’s success in cultivating customers that make different kinds of chips.
But just because Arm is increasing its foothold in data centers doesn’t necessarily mean it will be central to the future of AI. Among public companies, the main beneficiary of the AI boom has been Nvidia, because it makes chips that are specifically tailored to AI uses. Giddy investors pushed the chipmaker’s market cap above US$1 trillion for the first time in May after it surprised Wall Street with a bullish sales forecast that reflected the demand for these chips. Other companies that are more focused on general-use data centers, such as AMD and Intel, haven’t performed nearly as well.
Arm’s IPO is a chance to reach an audience that’s hungry for a story about profiting from the future of AI. But investors will get pickier as they form more nuanced theories about which companies will benefit from AI, according to Columbia Threadneedle Investment senior portfolio manager Rahul Narang. “One needs to be wary of pretenders versus contenders,” he says. “It’s important not to get caught in the hype.” - Bloomberg BusinessWeek
